IBM Engineering AI Hub CVE-2026-15322

Exploitation requires that IBM Engineering AI Hub emit a valid session token within a URL (query string or path) - the CWE-598 condition - and that the attacker gain access to a place where that URL is recorded or transmitted: reverse-proxy/load-balancer/web-server access logs, SIEM/log-aggregation stores, browser history on a shared machine, or the HTTP Referer header sent to an attacker-controlled or third-party external site. The limiting factor is that the vulnerability is passive: despite the CVSS PR:N/UI:N metrics, an attacker cannot request the tokens directly from the server and must obtain the leaked URL out-of-band, and the captured token must still be unexpired and un-rotated when replayed. Affected only on versions 1.0.0, 1.1.0, and 1.2.0.

The CVSS 3.1 base score is 7.5 (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N), reflecting network-reachable, unauthenticated, low-complexity access with high confidentiality impact and no integrity or availability impact - consistent with a pure information-disclosure/session-token leak. However, the CVSS vector overstates ease of exploitation for a CWE-598 issue: the vulnerability itself is passive, so an attacker cannot simply request the tokens on demand - they must first gain access to a location where the URLs are recorded (logs, proxy, referer, shared browser) or observe traffic. No EPSS score was provided in the input, the CVE is not in CISA KEV (not actively exploited), and no POC is referenced, so the realistic near-term risk is lower than the 7.5 suggests. This is a genuine but secondary priority: high-impact if a token is captured, but exploitation depends on secondary access to logging/referer data rather than a directly weaponizable network attack.

An attacker with read access to a shared reverse-proxy or web-server access log - or who operates an external site that receives the Referer header from an Engineering AI Hub user - extracts a valid session token embedded in a logged URL. The attacker replays that token against the application to hijack the victim's authenticated session and read the data that user can access. Given AV:N/AC:L, no special skill is needed once a token URL is obtained, but the attack hinges on first accessing that logged or referred URL rather than a direct server exploit.